Critical analysis of different Hilbert-Huang algorithms for pavement profile evaluation
Date
2010
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Pavement profile analysis is a major component in pavement
infrastructure management decision making for maintenance and rehabilitation. This
paper takes an in-depth look at pavement profile characterization and evaluation;
taking into account the inherent nature of road profile data i.e. non-stationary and non-
Gaussian. Although there have been several studies aimed at the analysis and
characterization of pavement profile, the bulk have been limited to applying relatively
conventional signal processing techniques, such as the Fourier analysis. Using this
approach, only the average condition of the local conditions can be represented; most
transient and changing signals will not be handled well due to the averaging effect of
the technique. The Hilbert-Huang transform operates at the scale of every oscillation,
an extremely important property for obtaining localized profile information. In this
work, the different algorithms of the Hilbert-Huang transform: Empirical Mode
Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD) and
Complex Empirical Mode Decomposition (CEMD) have been discussed and
implemented to extract useful information from road profile data. The robustness of
the algorithms is compared based on its ability to produce physically meaningful
Intrinsic Mode Functions (IMFs) which truly characterize the underlying process. The
results show that although all the methodologies yielded similar residual trends, the
CEMD produced physically meaningful and trusted IMFs whose information at the
various levels of decomposition could be used to extract profile information such as
the extent of deterioration and localized roughness information.